UMD Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/3
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a given thesis/dissertation in DRUM.
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
2 results
Search Results
Item Statistical Models and Optimization Algorithms for High-Dimensional Computer Vision Problems(2011) Mitra, Kaushik; Chellappa, Rama; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Data-driven and computational approaches are showing significant promise in solving several challenging problems in various fields such as bioinformatics, finance and many branches of engineering. In this dissertation, we explore the potential of these approaches, specifically statistical data models and optimization algorithms, for solving several challenging problems in computer vision. In doing so, we contribute to the literatures of both statistical data models and computer vision. In the context of statistical data models, we propose principled approaches for solving robust regression problems, both linear and kernel, and missing data matrix factorization problem. In computer vision, we propose statistically optimal and efficient algorithms for solving the remote face recognition and structure from motion (SfM) problems. The goal of robust regression is to estimate the functional relation between two variables from a given data set which might be contaminated with outliers. Under the reasonable assumption that there are fewer outliers than inliers in a data set, we formulate the robust linear regression problem as a sparse learning problem, which can be solved using efficient polynomial-time algorithms. We also provide sufficient conditions under which the proposed algorithms correctly solve the robust regression problem. We then extend our robust formulation to the case of kernel regression, specifically to propose a robust version for relevance vector machine (RVM) regression. Matrix factorization is used for finding a low-dimensional representation for data embedded in a high-dimensional space. Singular value decomposition is the standard algorithm for solving this problem. However, when the matrix has many missing elements this is a hard problem to solve. We formulate the missing data matrix factorization problem as a low-rank semidefinite programming problem (essentially a rank constrained SDP), which allows us to find accurate and efficient solutions for large-scale factorization problems. Face recognition from remotely acquired images is a challenging problem because of variations due to blur and illumination. Using the convolution model for blur, we show that the set of all images obtained by blurring a given image forms a convex set. We then use convex optimization techniques to find the distances between a given blurred (probe) image and the gallery images to find the best match. Further, using a low-dimensional linear subspace model for illumination variations, we extend our theory in a similar fashion to recognize blurred and poorly illuminated faces. Bundle adjustment is the final optimization step of the SfM problem where the goal is to obtain the 3-D structure of the observed scene and the camera parameters from multiple images of the scene. The traditional bundle adjustment algorithm, based on minimizing the l_2 norm of the image re-projection error, has cubic complexity in the number of unknowns. We propose an algorithm, based on minimizing the l_infinity norm of the re-projection error, that has quadratic complexity in the number of unknowns. This is achieved by reducing the large-scale optimization problem into many small scale sub-problems each of which can be solved using second-order cone programming.Item Quantum coherent phenomena in superconducting circuits and ultracold atoms(2010) Mitra, Kaushik; Lobb, Chris J; Physics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This thesis consists of theoretical studies of superconducting qubits, and trapped bosons and fermions at ultracold temperature. In superconducting qubits I analyze the resonant properties and decoherence behavior of dc SQUID phase qubits, in which one junction acts as a phase qubit and the rest of the device provides isolation from dissipation and noise in the bias lead. Typically qubit states in phase qubits are detected by tunneling it to the voltage state. I propose an alternate non-destructive readout mechanism which relies on the difference in the magnetic flux through the SQUID loop due to state of the qubit. I also study decoherence effects in a dc SQUID phase qubit caused by the isolation circuit. When the frequency of the qubit is at least two times larger than the resonance frequency of the isolation circuit, I find that the decoherence time of the qubit is two orders of magnitude larger than the typical ohmic regime, where the frequency of the qubit is much smaller than the resonance frequency of the isolation circuit. This theory is extended to other similar superconducting quantum devices and has been applied to experiments from the group at the University of Maryland. I also demonstrate, theoretically, vacuum Rabi oscillations, analogous to circuit-QED, in superconducting qubits coupled to an environment with resonance. The result obtained gives an exact analytical expression for coherent oscillation of state between the system (the qubit) and the environment with resonance. Next I investigate ultracold atoms in harmonically confined optical lattices. They exhibit a `wedding cake structure' of alternating Mott shells with different number of bosons per site. In regions between the Mott shells, a superfluid phase emerges at low temperatures which at higher temperatures becomes a normal Bose liquid. Using finite-temperature quantum field theoretic techniques, I find analytically the properties of the superfluid, Bose liquid, and Mott insulating regions. This includes the finite temperature order parameter equation for the superfluid phase, excitation spectrum, Berezinskii-Kosterlitz-Thouless transition temperature and vortex-antivortex pair formation (in the two dimensional case), finite temperature compressibility and density - density correlation function. I also study interacting mixtures of ultracold bosonic and fermionic atoms in harmonically confined optical lattices. For a suitable choice of parameters I find emergence of superfluid and Fermi liquid (non-insulating) regions out of Bose-Mott and Fermi-band insulators, due to finite boson and fermion hopping. I also propose a possible experiment for the detection of superfluid and Fermi liquid shells through the use of Gauss-Laguerre and Gaussian beams followed by Bragg spectroscopy. Another area I explore is ultracold heteronuclear molecules such as KRb, RbCs and NaCs. I obtain the finite and zero-temperature phase diagram of bosons interacting via short range repulsive interactions and long-ranged isotropic dipolar interactions in two-dimensions. I build an analytical model for such systems that describes a first order quantum phase transition at zero temperature from a triangular crystalline phase (analogous to Wigner crystal phase of electrons) to superfluid phase. At finite temperature the crystalline phase melts, due to topological defects, to a hexatic phase where translational order is destroyed but hexagonal orientational order is preserved. Further temperature increase leads to the melting of the hexatic phase into a normal dipolar Bose liquid.